摘要
提出用多模块神经网络的方法求解空间3R机械手的逆运动学多解。通过几何分析,将关节空间划分为多个只有唯一逆运动学解的关节子空间,每个子空间均用3个单输出的BP神经网络训练和求解。通过仿真试验并与其他方法对比,表明该方法不仅可以准确地划分逆运动学解的取值范围,还可以快速求得高精度的逆运动学多解。
A multiple module neural network method was proposed to solve the inverse kinematics multiple solutions of space 3 R manipulators.The joint spaces were divided into several joint subspaces with unique inverse kinematics solution through geometric analysis. Each subspace was trained and solved by 3 single output BP neural networks. After the simulation and comparison with other methods, results show that the method herein may accurately divide the value ranges of inverse kinematics solutions, and quickly obtain high precision inverse kinematics multiple solutions.
作者
肖帆
李光
游雨龙
XIAO Fan;LI Guang;YOU Yulong(College of Mechanical Engineering,Hunan University of Technology, Zhuzhou, Hunan,412007)
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2019年第10期1233-1238,共6页
China Mechanical Engineering
基金
湖南省自然科学基金资助项目(2018JJ4079)
关键词
机器人逆运动学多解
BP神经网络
关节子空间
3R机械手
inverse kinematics multiple solution of robot
BP neural network
joint subspace
3R manipulator
